Mads, thanks a lot for first round.
http://codereview.chromium.org/945002/diff/1/2 File src/ia32/stub-cache-ia32.cc (right): http://codereview.chromium.org/945002/diff/1/2#newcode1304 src/ia32/stub-cache-ia32.cc:1304: const int allocation_delta = 4; On 2010/03/15 08:51:34, Mads Ager wrote:
Please follow Google C++ style naming conventions: kAllocationDelta.
Done. http://codereview.chromium.org/945002/diff/1/2#newcode1317 src/ia32/stub-cache-ia32.cc:1317: // We fit. Don't forget to restore edx. On 2010/03/15 08:51:34, Mads Ager wrote:
Could you move the restore comment to where you actually restore edx
and
explicitly state that edx needs to hold the receiver when pushing the
element. Done. http://codereview.chromium.org/945002/diff/1/2#newcode1323 src/ia32/stub-cache-ia32.cc:1323: Immediate(Factory::undefined_value())); On 2010/03/15 08:51:34, Mads Ager wrote:
Shouldn't you fill the rest of the backing store with holes?
I only grow the elements by kAllocationData and the delta seems to be filled properly. I don't want to code memset in asm so I'd switch to more common 1.5x + 16 policy when/if I enable not initialized parts of fixed arrays. http://codereview.chromium.org/945002 -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev
